The prediction engine

Automatic model selection

n-Infinite automatically selects the best-fit modeling approach for each question in your survey, choosing from a set of proven machine-learning techniques: decision trees, random forests, regression models, gradient boosting (CatBoost), and TabPFN — a modern pretrained model well suited to smaller datasets.

You don't need to choose a model yourself; the platform evaluates each question and picks an appropriate method automatically, with manual override available for advanced users — every question row in the project detail has its own model dropdown:

Smart Match

Smart Match is an optional setting that steers predictions toward the most common ("majority") real-world answer for a given question, rather than a raw probability-weighted guess. It is most useful when your train and predict populations are known to be similar and you want the synthetic data to track real-world central tendencies closely.

Automatic recommendation. n-Infinite compares the distribution of your train and predict populations — weighting key demographic fields like age, gender and location more heavily — and recommends enabling Smart Match when the two populations look statistically alike. The recommendation badge appears in the project detail header, and the toggle sits in the bottom bar: